Category
383 papers
This paper argues that a small win does not show that routing did anything, the authors' or anyone else's, and argues that a small win does not show that routing did anything, theirs or anyone else's.
"List as many words as you can that start with M." The verbal fluency (VF) task is simple, yet even a typical university student only manages to produce about 15 words within 1 min, and there is substantial variability around this mean. The present study examined how linguistic and domain-general abilities contributed to VF performance in a large sample of healthy adult native speakers of Dutch (N = 571). We assessed the effects of linguistic knowledge, processing speed, short-term/working memory, and fluid intelligence on performance in the VF task. To examine whether linguistic and domain-general abilities contribute differently across VF task types, we included semantic trials (category-based generation: animals, food) and phonemic trials (letter-based generation: words beginning with S or M). We assessed the total number of correct words produced and the time to first response. Mixed-effects modeling showed that linguistic knowledge predicted the total number of correct responses in both semantic and phonemic VF. Short-term/working memory and processing speed were also significant predictors, but with smaller estimated effect sizes. Time to first response showed little effect of linguistic skills. We discuss how linguistic knowledge shapes the structure of the mental lexicon such that it affects both meaning-driven and form-driven access to lexical items. In addition, we provide updated norms for VF performance in Dutch and practical suggestions for using the task. (PsycInfo Database Record (c) 2026 APA, all rights reserved).
The results suggest that LLM data mixing should be treated not only as a prediction problem, but also as an experimental-design problem in which the proxy mixtures themselves can be chosen to improve statistical efficiency.
Agent harnesses record a failed tool call and its error message in the transcript and ask the model to continue, on the assumption that the error is corrective information, and it is found the gain is negative for every instruction-tuned model tested.
Reach audiences
Advertise in front of researchers, engineers, and readers.
This work introduces a general steering technique called Semantic Overlays: small learned adapters applied at chosen prefill positions to a frozen model's residual stream that defends against the broad class of prompt injections that add instructions in untrusted context.
This work introduces Mixture of Channel Experts (MoCE), a structured sparse channel-mixing layer, inspired by MoE, that replaces pointwise (1x1) channel-reduction projections and matches or exceeds dense baselines and prior channel-selection methods while reducing MACs by 16.7% and end-to-end latency.
This work presents MARS (Multi-Agent Relay of Specialized LLMs), a prompt-only framework in which each agent is a topic specialist---dynamic programming, graphs, strings, geometry, and so on---grounded by retrieval-augmented generation over an algorithm-theory corpus.
A reproducible reliability audit of the developer-accessible on-device foundation model is presented, framed as an oversight question: can a user or a resource-constrained developer tell when the model is wrong?
Stochastic analysis sharpens rather than erodes the thesis: the ignition boundary acquires a predicted width, and noise punishes the reactive policy that parks the system against it.
Results show that a small box-level module can reconcile question understanding with precise localization without retraining either backbone, and introduce RefineRank, which closes this gap at the candidate-box level.
Discrete diffusion provides an effective alternative to autoregressive radiology report generation by enabling iterative, bidirectional report refinement.
Whether off-the-shelf Large Language Models (LLMs) can effectively reason about taint flows in Android apps is investigated, and preliminary findings suggest that LLM reasoning may effectively complement traditional static taint analysis.
Data Management
What if pathology foundation models could do more with less? GigaPath-Flash and GigaTIME-Flash cut computational demands while maintaining strong performance, opening the door to larger studies and broader exploration. The post GigaPath-Flash and GigaTIME-Flash: Toward population-scale discovery with efficient pathology foundation models appeared first on Microsoft Research.
With millions of users across the world, Julia has been used to conduct cutting-edge research and to design new drugs, jet engines, heat pumps, and more.
Earth AI
Health & Bioscience